View
0
Download
0
Category
Preview:
Citation preview
In Partnership With You!
Update on Analysis of Shared Syndromic Data for Severe Respiratory Injury Surveillance (Internal Use Only)
Michael A. Coletta, MPH, Aaron Kite-Powell, MS
In Partnership With You!
2
Outbreak of electronic-cigarette, or vaping, product use-associated lung injury (EVALI) As of November 14, 2019: 49 states, Puerto Rico, District of Columbia (DC) and the U.S. Virgin Islands 2,172 confirmed and probable cases reported 42 deaths
As of October 15: In 3 months before symptoms, 11% reported exclusive nicotine use and
34% reported exclusive tetrahydrocannabinol (THC) use 86% reported any (THC)-containing products, 64% any nicotine
Timeline
9/20/2019 NSSP shares work on state epi call
7/10/2019 Wisconsin Department of
Health Services (WDHS) identifies 5 adolescents admitted in previous 30
days
8/15/2019 NSSP begins
engagement with syndromic community
of practice
8/23/2019Community Slack channel created for collaboration
on queries
8/29/2019 Query for e-cigarette
vaping product use added to CCDD searches in
ESSENCE
9/06/2019 IDPH published NEJM article
9/12/2019NSSP began
approaching sites for collaborative project with full data sharing
9/04/2019 CDC shares definitions of e-
cigarette or vaping use query and lung injury query
developed with IDPH
11/18/2019Preliminary analyses
to share
4
Queries E-cigarette or vaping: Terms related to e-cigarette use or vaping in the
chief complaint
Lung injury query, version 1 (sensitive): Includes acute respiratory illness, and excludes unrelated infectious disease, chronic disease, and injury; originally developed by IDPH for case-finding
Lung injury query, version 2 (specific): Includes only CDC-recommended codes for EVALI with same exclusions as version 1
NSSP ESSENCE Use: Total number of queries for lung injuries or e-cigarette or vaping
Aug. 29: E-cigarette and vaping query added to ESSENCE2,000
1,500
1,000
500
0
3/30 4/14 4/29 5/14 5/29 6/13 6/28 7/13 7/28 8/12 8/27 9/11 9/28 10/11
E-cigarette or vaping in chief complaint by age 2017–2019Percent of all ED visits for each age group
Ages 10-19
Ages 20-29
0.04
0.02
0
Percent of all ED visits for each sex
0.0150.015
0.010
0.005
0
E-cigarette or vaping in chief complaint by sex 2017–2019
8
Why share data? Without data sharing from sites, CDC cannot access details like patient age Response asked NSSP to help answer the question of when did this start?
– More sophisticated methods to explore– Three main approaches pursued
1. Difference of differences (ANCOVA)2. Text-based analyses3. Spatial and spatio-temporal analyses
Initially we used a time series approach to see if there was a any trend. We also used a difference in differences (ANCOVA) approach to determine the significance of the difference
between men and women between two time periods.
At a national level, large amount of seasonality within the data that is not controlled for with the denominator. However there is a statistically significant difference between men and women across both time periods.
In Illinois, we see the data lacks a strong trend for men prior to April 2019, and has a steep upward trend after that point. There is also a statistically significant difference between men and women across both time periods.
Methods
Findings
Trend analysis by sexObservation
At a national level we noticed a difference between men and women in the trend for severe respiratory illness. We saw this same trend in Illinois after they shared their data.
1st Analytic Approach - ANCOVA
Illinois lung injury syndrome (sensitive) by sex, ages 14–30 not discharged home
Before April 2019: males ages 14–30 seem to have flat trend
Post April 2019: males ages 14–30 seem to have increasing trend
Percent of visits
2017 2018 2019
1.5
1.0
0.5
0.0
Shared site trends for lung injury syndrome (sensitive) by sex among ages 14–30, not discharged home
Post September 2018: increase in SRI visits in ages 14–30
Post August 2019: increasing trend in both males and females ages 14–30
2017 2018 2019
0.6
0.4
0.2
0.0
Percent of visits
Lung injury syndrome (sensitive) – exploratory text analysis of shared data SRI Query (CCDD free text; Negations across multiple fields) limited to the
sites that shared data Methods for Exploratory Text Analysis
– Text and ICD Code N-gram Frequencies– Word network co-occurrence graphs– N-gram trends over time, by gender– Comparison of control and test time period
• Purpose is to answer: What N-grams are more likely to come from the current outbreak period as compared with control periods?
2nd Analytic Approach – Text Based Methods
Lung injury syndrome (sensitive) –weighted log odds ratio of top 10 discharge diagnosis unigrams
Tidylo R package: https://github.com/juliasilge/tidylo
Hypoxia
Dyspnea
Bronchopneumonia
Abnormal Diagnostic Imaging
Other secondary pulmonary hypertension
Psychoactive substance use
Myalgia
Pleurodynia
Nicotine dependence
Systemic lupus erythematosus
Acute respiratory distress
Nausea with vomiting
Bronchitis/pneumonitis due to chemicals
Nicotine dependence, other tobacco prod
Lobar pneumonia
Poisoning by cannabis (derivatives)
Sepsis, unspecified organism
Myalgia
Unspecified misadventures during care
Lung transplant
Lung injury syndrome (sensitive) –weighted log odds ratio of top 10 chief complaint bigrams
Tidylo R package: https://github.com/juliasilge/tidylo
Trend for top 10 bigram chief complaint terms Trend is as a percent of the lung injury (sensitive) query
Lung injury syndrome (sensitive) –weighted log odds ratio of top 10 chief complaint trigrams
Trend for top 10 trigram chief complaint terms
Lung injury query – observations
Chief complaints and ICD codes considered important in the test period included:– Chief Complaints: Shortness of breath, nausea/vomiting, cough
shortness, low O2, bilateral pneumonia, fever sepsis– Discharge Diagnoses: acute resp distress, nausea and vomiting,
bronchitis/pneumonitis due to chemicals, nicotine dependence (other products), lobar pneumonia, poisoning by cannabis and combinations
Anecdotally, these are similar to diagnostic codes of actual cases
New lung injury (specific), ages 11-34 not discharged home
0.009
0.006
0.003
0
Percent of visits
Lung injury query spatial analyses Methods and results are still in progress
– Standard Morbidity Ratio (SMR) analysis.• Analyzing region counts / region population.
– Pure Spatial analysis using SaTScan.• Analyzing geographical distribution of SRI visits.
– Spatio-Temporal ( Space-Time) analysis using SaTScan.• Analyzing geographical distribution within selected time
parameters.
3rd Analytic Approach – Spatial Methods
Conclusions ED visits with e-cigarette and vaping in the chief complaint began
increasing among ages 10-19 before 2019; may be general health effects rather than signal of EVALI outbreak
Text analysis of sensitive query for lung injury shows increase in symptoms such as nausea and vomiting associated with the EVALI outbreak in 2019
Specific query for lung increase increased beginning in June 2019
Limitations Records returned by query for e-cigarette and vaping in the chief
complaint not limited to lung injuries and likely strongly influenced by public awareness
Sensitive SRI query includes unexplained pneumonia, which cannot be separated from influenza-like illness; specific version has small numbers
Change in facilities sending data and quality of data sent may affect trend
Next Steps New England Journal of Medicine article about use of syndromic data
Continuing to refine and share text mining methods
Future spikes in trend or geographic clusters may be early signal of new risks, such as increasing exposure to dangerous e-cigarette or vaping products
For more information, contact CDC1-800-CDC-INFO (232-4636)TTY: 1-888-232-6348 www.cdc.gov
The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.
Recommended